Skip to main content

Full text of "Weiss Oscillations in Surface Acoustic Wave Propagation"

See other formats


Weiss oscillations in surface acoustic wave propagation 



Y. Levinson 

Condensed Matter Physics, The Weizmann Institute of Science, Rehovot 76100, Israel 

O. Entin-Wohlman 

School of Physics and Astronomy, Raymond and Beverly Sackler Faculty of Exact Sciences, 
Tel Aviv University, Tel Aviv 69978, Israel 



T — I 

Q 

m 



> 



o 
o 



A. D. Mirlin* and P. Wolfle 

Institut fiir Theorie der kondensierten Materie, Universitdt Karlsruhe, 76128 Karlsruhe, Germany 

(December 23, 1997) 

The interaction of a surface acoustic wave (SAW) with a a two-dimensional electron gas in a 
periodic electric potential and a classical magnetic field is considered. We calculate the attenuation 
of the SAW and its velocity change and show that these quantities exhibit Weiss oscillations. 

PACS numbers; 73.20.Dx, 73.50.Rb 



Oscillations in the dc magnetoresistance, known as 
Weiss oscillations, can be observed in high mobil- 

ity two-dimensional electron gas (2DEG) subject to a 
perpendicular magnetic field B and a periodic laterally- 
modulated potential. These oscillations are geometric in 
nature, occuring whenever the cyclotron radius Rc is an 
integral multiple of the modulation period, d. 

Surface acoustic waves (SAWs) are known to be a very 
effective tool to study the properties of a 2DEG [^-0. 
Recently the remarkable effect of a periodic potential on 
the SAW propagation has been observed in the quan- 
tum Hall regime ||]; this problem was addressed theo- 
retically in 1^. The case of a strong modulation, when 
the 2DEG is fragmented into quantum wires was exper- 
imentally studied in ||l^. In this Letter we focus on a 
different aspect: the possibility to observe Weiss oscilla- 
tions in the ac electric field of a SAW in weak (classical) 
magnetic fields. 

We show that when the SAW length is long compared 
to the grating period and the electron cyclotron radius 
the grating just renormalizes the conductivity <T(q, a;), 
where m and q are the frequency and the wave vector of 
the SAW. This means that the propagation of the SAW 
in the modulated 2DEG can be described as propaga- 
tion in a uniform 2DEG having an effective conductivity 
cre//(q, w) = (T(q, w) -|- (5(T(q, w), where the first term is 
the conductivity of the 2DEG without grating and the 
second is the renormalization due to the grating. The 
grating-induced contribution 5a will be shown to exhibit 
Weiss oscillations, as manifested in the SAW absorption 
and the velocity shift. 

Due to the piezoelectric properties of GaAs the defor- 
mation wave is followed by an electrical field wave, which 
in the plane of the 2DEG is E°(r,t) = E°exp(-iwt + 
iqr) -I- C.C.. We assume the field to be longitudinal, 
E" II q, which corresponds to the usual experimental 
geometry. This field is screened by the 2DEG, and 



because of the modulation the screened (total) field 
E(r, t) has spatial Fourier components with wave vec- 
tors qs = q + sp, (s — 0, ±1, ±2, ...), where p is the 
wave vector of the grating, p — 2-k / d. It follows that 
E(r, t) — Es exp(— iwt -|- iq^r) -I- c.c. The unscreened 
field has in these notations only the s = Fourier compo- 
nent, qo = q, and E° = 5s,oE°. The current density cre- 
ated by the SAW is j(r, t) = ^^is exp(—ia;t-|-iqsr) -I- c.c. 
and we define a matrix of tensorial conductivities as,s' by 

Since the SAW velocity v = to/q (equal to 2.8 x 
lO^cm/sec in GaAs) is much less than the light velocity 
one can neglect retardation effects and find the screen- 
ing field solving a quasistatic problem. The result is (a 
detailed account of our work will be given elsewhere | |ll| ) 



E, - E' 



27ri qU 
Lueo Qs 



E, 



(1) 



where is the effective dielectric constant of the back- 
ground. 

It follows from this equation that when E" || q, as 
assumed, then for all s one has Eg || q^, i.e. the total 
screened field is also longitudinal. In this case one can put 
Es = {c[s/Qs)Es and eliminate the tensorial properties, 
obtaining from Eq. (^ the relation between the screened 
and unscreened fields in a scalar form J2s' ^s,s'Esi = 
Ss.qE'^ , where the longitudinal dielectric constant matrix 
is es,s' — Ss,s' + i'2TTi/Lueo)<ls<7s,s' and the longitudinal 
conductivity matrix is = (qs/gs) • o-^y ■ (qs' /?«')• 

The absorption Q of the SAW is given by a spa- 
tial and temporal average of j • E. Using the field and 
current representations one finds for longitudinal fields 
Q = J2s s' Elas,s'Es' +C.C.. Expressing the screened field 
E in terms of the SAW field E° one has 

Q/|E0|2 = iaMie-')o,o + c.c. - -2(7M3(e"')o,o. (2) 

Here (e~^)o,o is the s = 0, s' = matrix element of 



1 



the matrix inverse to e and ctm = eov/2'K (equal to 
1 X 10-2 (eV/i) for GaAs). 

Furthermore, it can be shown that in the lowest 
order of the piezoelectric coupling constant a (in GaAs 
0^/2 — 3.2 X 10"'*) the renormalizcd dispersion relation 
of the SAW takes the form 



u^vq [l + (aV2)(e"')o,o] 



(3) 



As a consequence, both the relative shift of the SAW ve- 
locity Av/v and its attenuation coefficient per unit length 
r are given in terms of (e-^)o,o as follows 



Av/v = (aV2)3fJ(e"^)o,o 
r=-g(aV2)3(e-i)o,o. 



(4) 
(5) 



For the calculation of the longitudinal conductivities 
which define (e^^)o,o we start with the kinetic equation 
for the distribution function /(r, k, t) of the 2DEG elec- 
trons in a form used in 112(1 



9 „ e ^ d 
-+VV + -VXB- 



(eE(r,i)-VC/(r)) 
-if-{f))/r. 



d_ 



f 

(6) 



Here, v = k/m, U{r) denotes the modulating potential 
in the plane of the 2DEG, E(r,t) is the total electric 
field of the SAW, r is the momentum relaxation time 
(assuming short range scattering potential), the angular 
brackets (. . .) denote an average over the directions of k, 
and e = — |e| is the electron charge. 

When E(r, i) = the 2DEG is in equilibrium and the 
solution of Eq. (|) is/o(r,k) = fT{ek+U{r)), wheresk = 

/2m and fri^) is the Fermi distribution. To linearize 
Eq. (H) with respect to E = E(r) exp{—iujt) + c.c. we 
write /(r,k,t) = /o(r, k) -I- 5/(r, k, i) with 5/(r,k,i) = 
exp(-iu;i)(-^/o(r,k))F(r, k)-f c.c. 

For our purposes, it is sufficient to consider the zero 
temperature case, for which (— ^/o(r,k)) = 6{ek + 
U{r) — ep), where ep = \mv\ is the Fermi energy. 
Consequently the magnitude of the electron velocity 
v[t) is defined by the grating potential according to 
imt;(r)2 + U{y) = ep and F(r,k) becomes a function 
F(r, n) of the unit vector n = v/v{v). The current den- 
sity (taking into account the two spin orientations) is 
j(r) = e(m/7r)i;(r)(ni^(r, n)). 

After linearization with respect to the electric field E 
one obtains a linear equation for F , 



LF{r,n) = u(r)eE(r)n, 
where the operator L is 



(7) 



L = -iuj +-+ w(r)nV + (eVu(r) +lUc)-^ [ 

T 0(p T J In 

Here the angle Lp defines the direction of n, lOc — 
\e\B/mc is the cyclotron frequency and e = n x 



B/B. To calculate the tensorial conductivity ag'.s 
we put E(r) = Eg exp(«qsr) and represent F(r, n) — 
J2s' Xs',s(n)exp(iqs'r). 

Consider now a periodic potential U{r) = J7ocos(pr). 
In the case of weak modulation ry = J/o/ef ^ 1 and one 
can perform a systematic expansion of Eq. (^ in rj. This 
implies expanding the functions v{r) and Xs',s{i^) and 
comparing in the resulting equation terms with the same 
spatial dependence and of the same order in 77. One then 
finds that since the grating has only the first harmonic 
in p, the non- vanishing components of the tensorial con- 
ductivity are: (s,s) to order 77*^, (s ± l,s) to order 77^, 
and (s, s), (s ± 2, s) to order 77^. 

The explicit expressions for the longitudinal conduc- 
tivities are 



(0) n ^ 

(1) w 



lOJT 



(Gs) 



lVFT{Gs±ldf^Gs) 



2vpqsqs±i 
2 



(8) 
(9) 



V CTo 



(ivpTy 



X {Gsd;l,Gs+idt^Gs + Gsdt\Gs-id;^Gs). (10) 

(As is shown below, (e^^)o,o does not require the (s±2, s) 
elements of ct^^'. To simplify the results the Fermi veloc- 
ity is renormalizcd according to vp vp(l — 77^/16)). 

Here cto = {'m/2Tr)e'^VpT is the dc conductivity of a 
homogeneous 2DEG when B = and the following op- 
erators are introduced: d^' — nqs + kped/d(p , 



Gs 



ILOT + lVpTnC{, 



d 

+ UJcT — 

dip 



dip 
2^ 



(11) 



The explicit calculation of the conductivity can be per- 
formed in terms of the integral representation of the op- 
erator Gs given by 



Gsgiip) = Rsfjiv) + T^^^Rs, 



(12) 



l-(^.)' 

where the operator Rs is defined by 

R.9i^)=l r dip'e''^^'^''^'^g{ip'), (13) 



with Ks{p,p') = —v{p — p') — izs{smip — smip') and 
V — (r^"'" — iuj)/ujc: 7 = 1/i^cT and Zg — qsVp/uic- The 
angles ip and ip' are counted from the direction of and 
Rs, {Rs) stand for Rs\ and (Rsl), respectively. 

In what follows we consider a " fast" grating with a pe- 
riod shorter than the SAW length A, i.e. q p. (As an 
example, for lo / 2ti — 300MHz one has A — 9/xm while d 
varies from 0.1/ini to l^m). In this case one can present 
the result of inverting the matrix e in the following way 



2 



1 



(14) 



with o'e//(q, Lo) = a{q, to) +6a{q, w), where a{q, lu) is the 
longitudinal conductivity of a homogeneous 2DEG cor- 
responding to wave vector q and frequency lu and 



- „(2) 



,(1)^(1) 



'0,0 



1,1 



i{qo/qi)(TM 



,(0) 
1,1 



i{qo/qi)e^M 



(15) 



(Here I = —1). The result ( p^ ) for (5(t is valid up to the 
second order in 77, assuming 77 to be small. We will show 
below that nonetheless in the case qJ-P and for u>cT 3> 1 
Sa can be comparable to cr, since Sa/a oc 77^ (wct)^. Eq. 
( |l5| ) is however correct even in this case, since the higher 
order terms are small, of order 77^, compared to da. 

Note that for a fast grating we have approximately 
qi = qi = p and (7^1 = a["l = cr(p,cj), while aj^l = 
a{q,uj). As a result one can simplify Eq. ( |l5| ) to obtain 



(1) (1) , (1) (1) 
^0,1^1,0 + <I^I,o 

(t(p,w) - i{q/p)(7M' 



(16) 



The conductivities of a homogeneous 2DEG are given 

by [pl 



a(g„ w) EE =2ao 



UJ 



1 



iujT 1 — (Rs) 



where 



(Rs) = {n^/ smhniy)Jit,{zs)J-iuizs) 



, (17) 



(18) 



and J±ii, are the Bessel functions. 

We assume in addition that A is longer than the cy- 
clotron radius Rc = vf/oJc, i.e. zo = qvp/^c = qRc ^ 1- 
(For B — O.IT and vp = 1-3 x 10'' cm/sec the cyclotron 
radius Rc — 0.5/im, and for uj/2tt — 300MHz one finds 
qRc = 0.3). Since the Weiss oscillations take place at 
pRc ~ irn with n = 1, 2, . . ., the above assumption p ^ q 
implies that a considerable number of oscillations belongs 
to the region qRc <^ 1. 

In the case qRc "C 1 one finds from Eq. (^7|) the lon- 
gitudinal conductivity of the homogeneous 2DEG to be 



a{q,uj) ^ (To- 



1 — iujT 



(19) 



(1 — iojr)^ + {lOct)"^ + i{qvFTY' I'^'-OT^ 
while the correction (5cr, Eq. (|T^), is reduced to the form 



5(7(q,Lj) = 77 (To 



8(l-(i?o))2 



(20) 



The functions $1^2 depend on 7, t^, z = pRc = 'luRd d 
and on the propagation direction of the SAW given by 
the angle Q between q and p, 



$1=2 



$2 = - 



7 

(7/2 -\- 1)2 
77/ 



z(Ri) 

1 - {Ri) 
,,2 



DM. 
2z^{R{) 



7.2 + 1 (7.2 + 1)2 1 „ (^^) 



DM. 



where 



DM 



-^sin^i 



(21) 
(22) 

(23) 



From Eqs. (|T]), (|2|) and (H) it follows that when q ± p 
the grating induced part of the conductivity contains in 
addition to the small factor rf' also a factor [lOctY which 
can be large. 

The function ^ is independent of the propagation di- 
rection 



= (1 - iuJT) 



{Ri) , l+paB/2 



l-(i?i) 



(24) 



where as — eo/me^ is the Bohr radius (equal 100 A for 
GaAs). Using Eq. ( |l8|) to calculate (Rq) and in the 
functions $1^2 and ^ one can find (5cr(q, w) and the SAW 
propagation properties from Eqs. (^ and 

The Weiss oscillations are expected to be visible when 
LUcT > 1 and z = pRc > 1. (For B = O.IT and 
T =150 ps one finds lUcT — 40). We assume in addi- 
tion that ujT < 1. (For w/27r = 300MHz with t =150 ps 
one finds ujt = 0.3). With these approximations (when 
7 = 7.= (uJct)^^ <C 1) the results are more transparent. 

The longitudinal conductivity with no grating is given 
now by 



cr(g,cj) 



c^o 



(LOcTy 



i{qRc 



2UJT 



(25) 



Note that ao/{ujcT)'^ is the "rfc conductivity" at strong 
magnetic fields, which is obtained upon solving the ki- 
netic equation with an electric field constant in time and 
homogeneous in space. This " dc conductivity" corre- 
sponds to the condition {qRcY /2ujt -C 1, which is not 
always satisfied in SAW measurements. 

In the assumed range of parameters, the term ^$1 in 
Eq. ( po| ) can be neglected compared to $2, yielding 



(5CT(q,cj) = a, 



(2) 
0,0 



SiUcT^ 



l + i 



{qRcf 



2UJT 



X {-1 + (j){z) [cos^ e - [ujcTf sin^ 6] }. (26) 



with 



(z)=2zVo2(z)/[l-Jo2(z)]. 



(27) 



These results demonstrate that the effect of the grating 
and hence also the amplitude of the Weiss oscillations is 
stronger by a factor of {lUctY when the SAW propagates 
perpendicular to the grating, E° || q ± p, compared to 



3 



the case of parallel propagation, E° || q || p. This is be- 
cause in the "perpendicular" geometry the grating affects 
the Hall current, which is stronger. In the limit g ^ Eq. 
( p6| ) reduces to the result for the dc magnetoconductivity 
derived in p^ . 

2R,/d 

43 2 1 

— mil I' I 'I I — ' — ' — r — I — ■ — ■ — ' — I — ■ — 




B( Tesia) 



which can be large even for small rj. This is demonstrated 
in the figures, where rj = 0.01 is sufficient to produce 
strong oscillations {uJcT is as large as 100 at i? = 0.25T). 
For the parallel orientation q || p (6' = 0) the factor of 
(wct)^ is absent in ( p6| ) and the amplitude of oscillations 
is proportional to rj^ and thus small for weak modulation. 

Let us mention that in high mobility samples the im- 
purity potential is smoothly varying and the scattering 
probability is not isotropic, but rather peaked in the for- 
ward direction. This will strongly affect the damping of 
the oscillations with high oscillation number 2Rc/d ^ 1, 
but will not be very important for the first few oscilla- 
tions with 2Rc/d ~ 1. 

We thank A. Stern and F. von Oppen for discussions 
of unpublished results. This work was supported by the 
Alexander von Humboldt Foundation (Y. L.), the Albert 
Einstein Minerva Center for Theoretical Physics (O. E.- 
W.), the SEE 195 der Deutschen Forschungsgemeinschaft 
(A. D. M. and P. W.), and the German-Israeli Founda- 
tion. 



FIG. 1. Absorption coefficient F of SAW as a function of 
the magnetic field for the transverse orientation, p^q and the 



u} = 277 ■ 300MHz, T = 150ps, d = 300nm. The strength of 
the grating rj is equal to (from the top to the bottom): (no 
grating), 0.01, 0.02, 0.03, and 0.05. 




0.2 0.4 
B( TesIa) 



FIG. 2. Velocity shift Av/v of SAW as a function of the 
magnetic field. The values of all the parameters are the same 
as in Fig.l. 

We have evaluated numerically the attenuation and the 
velocity change according to Eqs. (^), (||), (p^), and 
( p0| ) using typical experimental parameters in the regime 
d, Rc ^ A. The results are shown in Figs. 1 and 2 for the 
orientation qJ-p. In this case the oscillations amplitude 
is proportional to (wct)^?7^ (see Eq. (26) for ~ 7r/2), 



[1 
[2: 

[3; 
[4: 

[6: 
[7: 

po 

[11 

[12 
[13 



Also at St. Petersburg Nuclear Physics Institute, 188350 
St. Petersburg, Russia. 

D. Weiss, K. v. Klitzing, K. Ploog, and G. Weimann, 
Europhys. Lett. 8, 179 (1989). 

R. W. Winkler, J. P. Kotthaus, and K. Ploog, Phys. Rev. 
Lett. 62, 1177 (1989). 

P. H. Beton, E. S. Alves, P. C. Main, L. Eaves, M. W. Del- 
low, M. Henini, O. H. Hughes, S. P. Beaumont, and 
C. D. W. Wilkinson, Phys. Rev. B 42, 9229 (1990). 
Y. Paltiel, U. Meirav, D. Mahalu, and H. Shtrikman, 
Phys. Rev. B 56, 6416 (1997). 

A. Wixforth, J. Scriba, M. Wassermeir, J. P. Kotthaus, 
G. Weimann, and W. Schlapp, Phys. Rev. B 40, 7874 
(1989). 

R. L. Willet, M. A. Paalanen, K. W. West, L. N. Pfeif- 
fer, and D. J. Bishop, Phys. Rev. Lett. 65, 112 (1990); 
R. L. Willet, R. R. Ruel, M. A. Paalanen, K. W. West, 
and L. N. PfeiflFer, Phys. Rev. B 47, 7344 (1993). 
R. L. Willett and L. N. Pfeiffer, Surface Science 
361/362, 38 (1996). 

R. L. Willett, K. W. West, and L. N. PfeiflFer, Phys. Rev. 
Lett. 78, 4478 (1997). 



F. von Oppen , A. Stern and B. I. Halperin, preprint cond 



mat/9712031 



G. R. Nash, S. J. Bending, M. Boero, P. Grambow, 
K. Eberl, and Y. Kershaw, Phys. Rev. 54, R8337 (1996). 
Y. Levinson, O. Entin-Wohlman, A. D. Mirlin, and 
P. Wolfle, to be published. 

C. W. J. Beenakker, Phys. Rev. Lett. 17., 2020 (1989). 
A. D. Mirlin and P. Wolfie, Phys. Rev. Lett. 78, 3717 
(1997). 



4